Journal of Agricultural and Resource Economics 42(3):350–371 ISSN 1068-5502Copyright 2017 Western Agricultural Economics Association

The Trade and Health Effects of Tobacco Regulations

Gregmar I. Galinato, Aaron Z. Olanie, and Jonathan K. Yoder

We examine how cross-country differences in tobacco regulations affect tobacco imports andconsumer health. We find that tobacco imports increase when a rich exporter’s tobacco regulationsare stringent relative to the regulations of its poor importing trade partner. The main policy drivermay be differences in marketing and counter-advertising tobacco regulations between tradingpartners. If a rich exporting country adopts counter-advertising tobacco regulations, mortalityand morbidity from tobacco-related diseases in the poor importing country increase by four andeighty smokers per million people annually, respectively. Our results highlight the importance ofaccounting for spillovers in an increasingly multilateral economy.

Key words: bilateral trade effect, gravity model, mortality, tobacco regulation

Introduction

The World Health Organization estimates that tobacco use kills 5 million people annually andprojects that world tobacco consumption will increase. Policy-makers have attempted to reducetobacco consumption using a variety of tobacco regulations—such as marketing bans that limitcigarette advertisements, age and smoking-location restrictions, and warning labels informingconsumers of the health effects of smoking (Laugesen and Meads, 1991).

The effects of these regulations on tobacco consumption have been analyzed extensively(Chaloupka and Warner, 2000). Marketing bans have yielded mixed results. In OECD countries,the effect of comprehensive marketing bans on tobacco consumption has ranged from insignificant(Stewart, 1993; Nelson, 2003) to modest (Laugesen and Meads, 1991; Saffer and Chaloupka, 2000).However, in developing countries, comprehensive marketing bans significantly reduce tobaccoconsumption (Blecher, 2008). There is a general consensus that smoke-free zones or smoking-location regulations significantly reduce cigarette consumption (Chapman et al., 1999). Similarly,counter-advertising efforts have been effective in reducing tobacco consumption in the United States(Lewit, Coate, and Grossman, 1981; Schneider, Klein, and Murphy, 1981; Warner, 1981; Baltagi andLevin, 1986; Hu, Sung, and Keeler, 1995).

While tobacco consumption is projected to increase in developing economies, tobaccoconsumption is declining in developed economies due, in part, to a variety of related regulations.Production in developed economies has not decreased as fast as domestic consumption sincedeveloped countries are a significant contributor in the export of tobacco (Food and AgricultureOrganization of the United Nations, 2003). The United States and European Union were twoof the top three exporters of raw tobacco from 1998 to 2000 (Commission of the EuropeanCommunities, 2003). There are assertions that tobacco industries in developed economies targetdeveloping countries with relatively more lax tobacco regulations (World Health Organization, 2008;Smith, 2015). Although extensive analysis on the impact of tobacco regulations on own-countrytobacco consumption has been conducted, no study has investigated the spillover effects of tobacco

Gregmar I. Galinato is an associate professor and Jonathan K. Yoder is a professor in the School of Economic Sciences atWashington State University. Aaron Z. Olanie is a Major Account Executive at Economic Modeling Specialists International.We thank Hayley Chouinard for early comments in conceptualizing the research idea and Ben Cowan for suggestions duringthe revision process.

Review coordinated by Tian Xia.

Galinato, Olanie, and Yoder Trade and Health Effects of Tobacco Regulations 351

regulations on tobacco trade flow. Given the projected rise in tobacco consumption in developingeconomies and the importance of trade in facilitating this rise, understanding the effects of tobacco-related regulations on tobacco trade flows is important to help develop policies that can further curbtobacco use.

This article fills this gap in the literature by examining the effects of the difference betweenimporter and exporter tobacco regulations on the flow of tobacco trade and its consequent healtheffects. We use a gravity model to test the effect of differences in tobacco regulations betweentrading partners on tobacco imports. The empirical results provide evidence for the extent towhich domestic regulations have spillover effects on other countries. We also simulate the spillovereffects of differences in tobacco regulations on mortality and morbidity from imported tobaccouse by combining our estimates with complementary results from studies the literature. Our maincontribution is to determine empirically if differences in tobacco regulations between bilateraltrading partners significantly affect tobacco trade, which would imply spillover effects fromunilateral changes in tobacco regulations. The presence of spillover effects implies the need forpolicies that increase the local and global effectiveness of tobacco regulations.

Several strands of literature have investigated the spillover effects of regulations on tradepatterns. In the context of food safety, Disdier, Fontagné, and Mimouni (2008) find that sanitary andphytosanitary regulations have no significant impact on agricultural trade between OECD countriesbut have a significant impact on developing countries, where regulations are more lax. In terms ofenvironmental pollution, the pollution haven hypothesis suggests that if a country has very strictregulations, firms’ environmental regulation compliance costs are high, giving firms an incentive torelocate to countries with less stringent environmental regulations (Grossman and Krueger, 1993,1995). The mechanism driving the impact of regulations on trade in these strands of literature issupply-oriented. However, the underlying mechanisms driving the effects of tobacco regulations ontrade stems from both supply- and demand-side factors.

Several studies have analyzed the effects of trade policies on tobacco consumption andproduction. Regional trade agreements significantly affect tobacco firms’ marketing, lobbying,and organizational strategies (Holden et al., 2010) as well as their market shares (Chaloupka andLaixuthai, 1996). Eliminating tobacco production subsidies in response to trade agreements initiallydecreases tobacco consumption, but tobacco production shifts to other countries and tobacco importssignificantly increase (Stoforos and Mergos, 2004). These changes in production, exports, andimports highlight the need to investigate any potential effect of tobacco regulations on trade flows.

To examine the trade and health implications of tobacco regulations, we incorporate fourcategories of tobacco regulations—marketing regulations, counter-advertising and educationmandates, smoking-location regulations, and age restrictions—into a gravity model. Marketingregulations restrict where and to what audience tobacco can be advertised. Tobacco counter-advertising includes health warnings or advertising campaigns designed to reduce smoking or,more broadly, any form of media or message contrary to the messages promoted by tobaccocompanies. Smoking-location regulations prohibit smoking in specific public areas. Age restrictionsare restrictions on the minimum age of tobacco-product purchasers. Tobacco marketing regulationsand counter-advertising programs affect demand for the entire market, while age and smoking-location regulations affect only specific consumer types in the market.

Differences in tobacco regulations between trading partners may either increase or decreasetobacco trade. Members of a trade group may have similar preferences regarding public healthprotection and therefore may adopt similar tobacco regulations, which in turn may lead tomore balanced tobacco trade. The European Union is one example, where member stateshave taken various tobacco control measures in the form of legislation, recommendations, andinformation campaigns such as advertising restrictions for tobacco products or anti-smokingcampaigns (Institution for European Environmental Policy, 2012).1 Alternatively, large differences

1 Even though EU member states adopt similar legislation, implementation may vary from country to country.

352 September 2017 Journal of Agricultural and Resource Economics

in regulations between trading partners could lead to an influx of goods to countries whereregulations are relatively lax since it is easier for some firms to penetrate markets with fewerregulations. This type of relationship has been studied in the pollution haven literature (Levinsonand Taylor, 2008). A similar relationship may occur with tobacco regulations, in which tobaccotrade flows away from developed and toward developing countries, where tobacco regulations areless stringent.

Our main variables of interest are cross-country differences in a set of tobacco regulationsbetween bilateral trading partners. Tobacco regulations are likely to be endogenous because demandfor imported tobacco could influence regulations affecting its consumption and suppliers of tobaccoare incentivized to influence trade conditions to their benefit. The relative position and lobbyingeffectiveness across countries could therefore impact the difference in tobacco regulations betweentrading partners.

We use several instruments to identify the effect of tobacco regulations on net imports, includinga measure of the susceptibility of the government to lobbying, average asbestos production,and an environmental sustainability index. Lobbying affects the marginal stringency of tobaccoregulation, which affects supply of tobacco regulations. Average asbestos production affects theperceived marginal damages from tobacco-related illnesses, which influences the demand fortobacco regulations. Finally, the environmental sustainability index is included to account for ageneral propensity for environmental and health regulation that is overall exogenous with respectto the tobacco market. None of the variables influence tobacco imports directly, only indirectlythrough tobacco regulations.2

The hypotheses we test elucidate the effects of differences in tobacco regulations on tobaccotrade and, consequently, consumer health. We find two important results. First, tobacco importsincrease when an exporter’s tobacco regulations are stringent relative to the tobacco regulationsof its importing trade partner and if the exporting country is rich and the importing country ispoor. The driver of this result may be the difference in tobacco marketing and counter-advertisingregulations between bilateral trading partners. When the exporting country is faced with constraintson advertising or is countered with vigorous anti-smoking campaigns, it may respond by exportingtobacco products to countries with lax marketing regulations and counter-advertising regulations.Second, the simulated spillover health effects can be significant. For example, given an incomegap of $25,000 per capita between trading partners, when an exporting country adopts counter-advertising tobacco regulations, the importing country experiences an increase in mortality fromtobacco-related cancer of four smokers per million people annually. This mortality rate is equivalentto the number of female deaths from alcohol use disorder worldwide (World Health Organization,2004).

Our results have important policy implications. Harmonizing tobacco regulations among tradingpartners, especially between rich exporters and poor importers, may increase their effectivenessin reducing tobacco imports, tobacco consumption, and health-related illnesses. Thus, our resultssupport the idea of lobbying to increase tobacco regulations in low- and middle-income countries.Doing so not only reduces consumption of domestic tobacco products but also prevents an influx ofinternational tobacco products from high-income countries.

Conceptual Framework

We present a model that yields an estimable empirical equation linking the effect of differences intobacco regulations on tobacco trade flows by modifying Anderson and van Wincoop’s 2003 gravitymodel. This model helps identify variables that need to be included in our empirical estimation andallows us to correctly interpret our results.

2 Since we use the difference in tobacco regulations in our regressions, we use the difference in government response tolobbying, the difference in average asbestos production, and the difference in environmental sustainability index betweenbilateral trading partners as instruments in our model.

Galinato, Olanie, and Yoder Trade and Health Effects of Tobacco Regulations 353

Tobacco Regulations and the Gravity Model

We assume n countries exist, each of which produces differentiated tobacco products that are eitherconsumed domestically or exported. The utility function for a representative consumer in country jis given by

(1) Ui j (ci j) =

[n

∑i=1

(ci j

βi

) σ−1σ

] σσ−1

,

where ci j is consumer j’s consumption of country i’s goods, βi is the share parameter for countryi’s good, and σ is the elasticity of substitution. The consumer’s budget constraint is y j = ∑

ni=1 pi jci j,

where y j is country j’s total income and pi j is the price country j’s consumers face for country i’sgoods. The consumer maximizes utility subject to a budget constraint.

The price of foreign goods takes the form pi j = piti j, where pi is the exporter’s supply price onthe world market and ti j is the per unit trade costs between country i and country j. The pricescountry j faces are different from the domestic prices in country i because they include tradecosts, defined as all costs (other than production costs) sustained in moving a good to its finalnondomestic use. Trade costs are directly affected by regulations and policy barriers, and they areindirectly affected by geographic and cultural variables through transportation costs (Gallup, Sachs,and Mellinger, 1999; Anderson and van Wincoop, 2004).

We incorporate the difference in tobacco-regulation stringency between trading partners intotrade costs by defining trade costs as

(2) ti j = eα(ri−r j)eCijBdbi j (1 + vi j) ,

where α is a parameter associated with differences in tobacco regulations between each bilateraltrade partner, ri is an aggregate tobacco-regulation index for the exporting country and r j is anaggregate tobacco-regulation index for the importing country, Cij is a row vector of cultural andgeographic characteristics, B is a column vector of associated cultural and geographic parameters,di j is the distance between countries i and j, b is a trade elasticity of distance, and vi j is country j’sad valorem import tariff on country i’s goods.

We also construct a disaggregate measure based on the difference in each tobacco regulationaffecting the market. We re-write equation (A2) as

(3) ti j = eβ

(rmci −rmc

j

)+γ

(rsai −rsa

j

)eCijBdb

i j (1 + vi j) ,

where rmc is an index of marketing and counter-advertising tobacco regulations, rsa is an index ofsmoking-location and age regulations, and β and γ are parametric weights associated with eachcountry pair for each regulation type.3

The solution to the model yields the following import demand equation weighted by the productof GDP between country trading pairs (see Appendix A for details):

(4)xi j

y jyi=

1yw

eα(ri−r j)eCijBdbi j (1 + vi j)

Pjπi

1−σ

,

where yw is world income, Pj ≡(

∑ni=1

yiyw

(eα(ri−r j)eCijBdb

i j (1 + vi j)/πi

)1−σ) 1

1−σ

is a measure

of inward multilateral resistance, and πi ≡(

∑ni=1

y jyw

(eα(ri−r j)eCijBdb

i j (1 + vi j)/Pj

)1−σ) 1

1−σ

is a

3 Ideally, we would disaggregate regulations further into four types, but since we only have two main instruments we canonly just identify the equation for two types of tobacco regulations. Disaggregating further would create an underidentifiedsystem.

354 September 2017 Journal of Agricultural and Resource Economics

measure of outward multilateral resistance. Multilateral resistance terms account for barriers to tradeaffecting both countries i and j in their trade with all trading partners and come in the form ofdiffering price indices (inward) or competition (outward) (Anderson and van Wincoop, 2003).

We find that tobacco regulations have two separate effects on a country’s tobacco import demand.The first is the direct effect, which is found in the term e(1−σ)α(ri−r j) in equation (4). The second isthe indirect effect, through the two multilateral resistance terms.

Empirical Model

Estimating equation (4) presents three challenges. First, our regulatory variables may be endogenousbecause a unilateral change in tobacco regulations by an importer in response to tobacco importswill alter the difference in tobacco-regulation stringency between bilateral trading pairs. We use aconceptual framework developed by Keohane, Revesz, and Stavins (1998) that classifies politicaleconomy determinants of regulations to identify potential instruments. Assume that t is the levelof tobacco-regulation stringency. Demand for tobacco regulations, tD, can be formulated as tD

(marginal benefit of tobacco regulations, public preferences) while supply of tobacco regulations, tS,can be formulated as tS (marginal cost of tobacco regulations, institutional factors). In equilibrium,the marginal value of tobacco regulations occurs when its marginal benefit equals its marginal cost.Here, both the tobacco-regulation stringency level and the marginal value of tobacco regulations areendogenous. The equilibrium level of tobacco-regulation stringency depends on the marginal valueof tobacco regulations, public preferences, and institutional factors. Thus, a reduced-form equationspecifying the determinants of tobacco-regulation stringency includes institutional factors and publicpreferences, which are exogenous supply- and demand-side factors, respectively.

We identify two main instruments—one is a proxy for institutional factors and the other isa proxy for public preferences. Neither of the factors directly influences tobacco imports. Thesupply of tobacco regulations is affected by lobbying and the political process. A government moresusceptible to lobbying contributions from a consumer lobbying group (tobacco manufacturer) islikely to establish stringent (lax) tobacco regulations. We use an index of corruption control as aproxy for government influenceability4 and calculate the difference in corruption control betweentrading partners as an instrument for the difference in tobacco regulations. Demand for tobacco-regulation stringency is affected by consumers’ perception of the marginal damage from tobacco use.Exposure to asbestos increases the marginal damages from tobacco (McFadden et al., 1986), therebychanging the perceived marginal damages from tobacco-related illnesses (Li et al., 1984). Asbestosexposure exacerbates tobacco-related illnesses, leading to an increase in perceived marginal damagesif the user attributes most of the illness to tobacco use. Alternatively, perceived marginal damagescan decrease if the user attributes the illness more to asbestos exposure. We use a measure ofasbestos production per capita as a proxy for asbestos exposure and determine the difference inasbestos production per capita between trading partners as an instrument for the difference in tobaccoregulations. Both of the main instruments are unlikely to affect tobacco imports directly but mayaffect them through the difference in tobacco regulations.

We also include an additional instrument that captures the effectiveness of environmentalregulations, which may correlate with how governments protect the health of the populace. Weuse the difference in environmental sustainability index between countries as an added instrumentto check robustness of our results.

Second, gravity equations are typically estimated using a log-linear specification. Using a datasetwith all possible trading partner combinations results in a large number of zero trade flows. Takingthe log of trade would create missing values for these observations and ignore a large amount ofimportant information. Two common estimation strategies involve either dropping observations

4 This is not to suggest that the political influence of lobby activity implies corruption. However, given availability of thesedata, it is not unreasonable to hypothesize that a corruption index is correlated with propensity to be influenced by interestgroup lobby effort. We retain the definition of corruption from the International Country Risk Guide.

Galinato, Olanie, and Yoder Trade and Health Effects of Tobacco Regulations 355

with zero trade or adding a small positive number to the dependent variable and taking its logand, consequently, estimating via ordinary least squares. Neither approach is desirable because bothproduce inconsistent parameter estimates.

Other empirical strategies have been developed to address the zeros problem.5 Santos Silva andTenreyro (2006) proposed a Poisson pseudo-maximum likelihood (PPML) estimator as a solution,in part because of its superior performance in the presence of heteroskedasticity.6 Using the PPMLestimator, we estimate our gravity equation in levels rather than the log-linearized form. Rewritingequation (4) in levels yields

(5)xi j

yiy j= exp(Zi jΘ) = exp

((1− σ)α (ri − r j) + (1− σ)CijB− lnyw

+(1− σ)b lndi j − (1− σ) lnPj − (1− σ) lnπi

),

where Zi j is a vector of covariates for each country pair and Θ is a vector of parameters. Theresulting Poisson log-likelihood function is

(6) lnL =n(n−1)

∑k

(−exp(Zi jΘ) +

xi j

yiy jZi jΘ − ln

xi j

yiy j!),

where n is the number of countries. The PPML estimator represented by equation (6) can becalculated by selecting the parameters that equate the following score equations equal to 0:

(7)n(n−1)

∑k

(xi j

yiy j− exp(Zi jΘ)

)zh

i j ∀ h = 1,2, . . . ,H,

where H is the number of covariates and zhi j is the hth covariate. As the score equations suggest, the

response data can include nonintegers and need not be Poisson distributed for consistency as long asthe conditional mean of trade is correctly specified as E

[xi j

yiy j

∣∣∣Zi j

]= exp(Zi jΘ) (Santos Silva and

Tenreyro, 2010).Finally, the multilateral resistance terms are not directly measurable from the data. Because

Anderson and van Wincoop (2004) were concerned with the parameter estimates of the multilateralresistance terms, they implemented a complex algorithm to simulate them. Rather than simulating,they can be proxied using a fixed effects approach by including country-specific dummies (Andersonand van Wincoop, 2003; Feenstra, 2004). This approach allows us to easily control for multilateralresistance—such as tariff and nontariff barriers—from other trading partners, but there are twotradeoffs. First, we are not able to measure the indirect effect of tobacco regulations through themultilateral resistance terms. Second, since we allow for country fixed effects, only variables specificto the bilateral trading pair can be included as regressors.

We test the following hypothesis on the effect of tobacco regulations on tobacco imports: Theeffect of aggregate tobacco regulation differences between the exporter and importer is to increasetobacco imports. Affirmation requires (1− σ)α > 0 in equation (5). In general, more stringenttobacco regulations in the exporting country are likely to lead to a reduction in domestic production(Chapman et al., 1999); instead, firms will sell to other jurisdictions with less stringent tobaccoregulations. One possible mechanism that allows this effect is that stringent tobacco regulations inexporting countries may free up resources that otherwise would have been spent in the domesticeconomy, allowing exporting firms to market in countries with less stringent regulations. For

5 Felbermayr and Kohler (2006) set up a Wooldridge (2002) style corner solutions model and used a Tobit regressionfor estimation. Using a Heckman-style selection process, Helpman, Melitz, and Rubinstein (2008) developed a two stageestimation procedure. However, both studies estimated a log-linear form of the gravity equation.

6 Using Monte Carlo simulations, the PPML estimator is favorable over the Tobit estimator when data are generated usinga constant elasticity model and the probability of observing zero is not independent of the regressors (Santos Silva andTenreyro, 2011).

356 September 2017 Journal of Agricultural and Resource Economics

example, there is evidence showing that advertising spending for UK tobacco firms did not changesignificantly after more stringent domestic tobacco regulations were implemented (Elliott, Wei, andLenton, 2010). If the regulations across exporters and importers are the same, there is no spillovereffect from the exporter’s regulations on tobacco imports. The policy implication is that matchingtobacco regulatory stringency between trading partners will reduce tobacco-related spillover effects.

Differences in counter-advertising and marketing regulations between countries may have adifferent effect on tobacco trade than differences in age or smoking-location regulations. Counter-advertising and marketing regulations have the potential to affect the entire population of currentand potential tobacco users, while age and smoking-location regulations affect only specific groupsof consumers. Difference in particular tobacco regulations between exporter and importer may alsoincrease trade instead of decreasing it. For instance, if member countries within a trade bloc requirespecific regulations before trade, we may see more trade occur as the difference in particular tobaccoregulations becomes smaller.

Data Description

We compile cross-sectional data from the year 2000,7 including tobacco import value, grossdomestic product (GDP), country-specific tobacco regulations, and bilateral distances and country-pair characteristics. Appendix B lists the data sources.

The trade data include total import values in millions USD of manufactured and unmanufacturedtobacco between two bilateral trading partners from the year 2000. There are 97 countries with dataon tobacco regulations in our sample, so we build a trade dataset including all possible tradingpartners, providing 97× 96 = 9,312 total observations.

We created four tobacco regulatory indices: a counter-advertising regulation index, a marketingregulation index, an age regulation index, and a smoking-location regulation index. Each indexwas calculated based on a country’s adherence to a subcategory (Appendix C). A country mayindicate either complete restriction, partial regulation, or no regulation for a particular subcategory.We assign a score of 1 when a country indicates complete restriction and a score of 0 for partial or noregulation. Then we add the scores for all subcategories and divide by total subcategories to arriveat an aggregate tobacco index. A country that adopts all subcategory criteria for a tobacco regulationhas the highest level of regulation, with an index of 1. In contrast, a country with zero regulationsadopted has an index of 0.8

Our main variable of interest is the difference between exporter and importer aggregate tobaccoindex, so we calculate the aggregate tobacco index for an exporter and subtract the aggregate indexfrom its importing trade partner. We also create two other tobacco regulation measures. One showsthe difference in marketing and counter-advertising regulations and the other is the difference insmoking-location and age regulations between exporter and importers. Both indices are created bytaking the average of the two relevant indices to arrive at an index between 0 and 1. We then subtractthe importer’s index from the exporter’s index.

Bilateral distances were calculated based on the shortest distances measured on the earth’ssurface, regardless of actual transportation routes. The cultural and geographic characteristicsinclude three dummy variables indicating whether the countries share a border, a common language,or a colonial link. We also include a trade treaty dummy in some empirical specifications in additionto our multilateral resistance dummies.

7 Ideally, we would compile a panel dataset to account for cross-country and time variation, but we are limited by ourtobacco regulation variables, which are not surveyed annually by the World Health Organization and show little variation, ifany, from when they were first collected in 2000. More importantly, since we do not have annual data for our instruments, wewould not be able to identify our endogenous tobacco regulations.

8 For example, there are four age regulation subcategories: sales to minors, age verification for sales, vending machinessales, and free tobacco products. If a country bans sales to minors, partially adheres to age verification for sales, allowsvending machine sales, and disallows free tobacco products, we assign a score of 1, 0, 0, and 1, respectively. Therefore, theresulting index for age regulation for that country is 0.5 given a sum of 2 out of a maximum of 4.

Galinato, Olanie, and Yoder Trade and Health Effects of Tobacco Regulations 357

The first instrumental variable is the difference in average asbestos production per capita betweentrading partners and is collected from the U.S. Geological Survey. Sixty-one countries had producedasbestos between 1930 and 1970. Historical population demographics from these countries weregathered to obtain asbestos production per capita for years where data is available, and the averagewas derived. The long lag in asbestos production per capita not only reduces any potential directeffect on tobacco imports during our sample time period, but the time frame was also chosen toavoid years in which asbestos regulations started in the 1970s. Other countries in our sample eitherdo not produce asbestos or produce a negligible amount and therefore do not report, so we assign avalue of 0.

The second instrumental variable is the difference in corruption control index from theInternational Country Risk Guide, which has been widely used since its introduction by Knackand Keefer (1995). The index measures the extent to which financial corruption (such as bribes forprotection) and insidious forms of corruption (such as lobbying to affect policy) are prevalent inthe country. The index captures instability of the political process as well as distortions in economiccompetition from nepotism or favoritism from those in power. Scores range from 0 to 6, where a highindex indicates more corruption control or less government influenceability by lobby groups.9 Wetake the average of the available corruption control index data from 1984 to 1995 and obtain seventy-eight country observations, then we take the difference in the corruption control index betweenimporter and exporter. If only seventy-eight countries are used, our total observations is reduced byone-third since we are left with 78× 77 = 6,006 observations. We employ multiple imputation usingtwenty imputations to fill in missing observations for the nineteen countries, assuming a normaldistribution.

Aside from our two main instruments, we also use the difference in environmental sustainabilityindices between country pairs from Yale University’s Center for Environmental Law and Policy. Thecomposite index ranges from 0 to 100 and is based on various measures that reflect pollution levels,natural resource endowments, and environmental regulations (World Economic Forum, 2001). Inour sample, Finland had the highest value while Saudi Arabia scored the lowest. Thirteen countriesfrom our sample do not have an index value, so we also use multiple imputation here to completeour sample.

Table 1 presents summary statistics of our data. The mean level of tobacco imports was $1.5million USD in 2000. For developed countries (those with per capita GDP above $4,085), themean tobacco import level is higher, at $2 million, while for developing countries (those withGDP per capita lower than $4,085), mean tobacco import is lower, at $0.25 million. Across ourfour measures of tobacco regulation indices, developed economies have more stringent tobaccoregulations than developing economies. Finally, the mean tobacco regulation difference betweenunique bilateral pairs is 0.215. The mean difference between unique bilateral pairs in marketing andcounter-advertising regulations is 0.289, slightly larger than the mean difference between uniquebilateral pairs in age and smoking-location regulations at 0.248.10

Results

We estimate model parameters for the full sample and include exporter/importer pairs with largeincome gaps. We compare the results with and without instrumental variable (IV) estimates using

9 The ratings are from subjective assessments by ICRG staffers based on available information from in-depth countryassessments.

10 The mean difference in regulations for all country pairs is always 0 since the regulation difference of one exporter-importer pair will be the negative value of the regulation difference when looking at the same two countries with the roles arereversed. This is why the mean differences of regulations in table 1 for all data are 0.

358 September 2017 Journal of Agricultural and Resource Economics

Table 1. Summary Statistics

All

Countrieswith GDPPer Capita≥≥≥ $$$444,,,000888555)

Countrieswith GDPPer Capita<<< $$$444,,,000888555)

Variable Mean St. Dev Min Max Mean MeanTobacco trade imports (inmillions USD)

1.528 31.022 0.000 2,539.341 2.046 0.251

Log of total GDP 3.501 2.066 −0.184 9.239 3.494 1.881

Contiguous 0.024 0.153 0.000 1.000

Common language 0.132 0.338 0.000 1.000

Common colony 0.018 0.135 0.000 1.000

Trade Treaty dummy 0.461 0.498 0.000 1.000

Log of trade distance 8.687 0.848 4.088 9.894

Counter-advertising regulations 0.340 0.474 0.000 1.000 0.362 0.286Marketing regulation index 0.330 0.253 0.000 1.000 0.384 0.196

Age regulation index 0.291 0.282 0.000 1.000 0.337 0.179

Smoking-location regulationindex

0.281 0.234 0.000 0.923 0.300 0.234

Tobacco regulation difference(exporter minus importer)

0.000 0.264 −0.702 0.702

Marketing andcounter-advertising regulationdifference (exporter minusimporter)

0.000 0.359 −0.875 0.875

Age and smoking-locationregulation difference (exporterminus importer)

0.000 0.306 −0.808 −0.808

Corruption control index 3.658 1.501 0.016 6.000 4.004 2.654

Average asbestos production percapita from 1930 to 1970 (metrictons per capita)

0.003 0.014 0.000 0.100 0.004 0.00002

Environmental sustainabilityindex

52.230 11.428 29.800 80.500 44.209 55.301

Notes: The mean tobacco regulation difference, marketing and counter-advertising regulation difference, and age and smoking-locationregulation difference for unique bilateral pairs are 0.215, 0.289, and 0.248, respectively.

PPML. We include imputed values for our instrumental variables in all specifications.11 We thencombine these estimates with existing parameters in the literature to simulate the health effects oftobacco consumption and examine the trade-mediated effects of tobacco regulations on health.

11 Appendix D shows the PPML and IV-PPML results without imputed values for the instruments. The sample size withoutimputed values is 2/3 of the sample with imputed values. The results of the regressions that exclude observations with imputedvalues for the instrument are similar in sign and significance with specifications (1) and (2) of the PPML regressions in table3 that include imputed values. Thus, if there is little to no endogeneity, results are consistent with and without imputed values.However, the significance is statistically weaker with IV PPML relative to specifications (1) and (2) of the IV PPML resultsin table 3. The loss in statistical power with the IV estimates is because the set of countries lost in the no-imputation caseare mostly developing countries, which leads to a loss of observations representing trade between developed and developingcountries that tend to have significant differences in tobacco regulations.

Galinato, Olanie, and Yoder Trade and Health Effects of Tobacco Regulations 359

Tabl

e2.

Det

erm

inan

tsof

Toba

cco

Reg

ulat

ions

:Fir

st-S

tage

Reg

ress

ion

Res

ults

Dep

ende

ntVa

riab

le

Toba

cco

Reg

ulat

ion

Diff

eren

ce

Mar

ketin

gan

dC

ount

er-

Adv

ertis

ing

Reg

ulat

ion

Diff

eren

ce

Age

and

Smok

ing-

Loc

atio

nR

egul

atio

nD

iffer

ence

Toba

cco

Reg

ulat

ion

Diff

eren

ce

Mar

ketin

gan

dC

ount

er-

Adv

ertis

ing

Reg

ulat

ion

Diff

eren

ce

Age

and

Smok

ing-

Loc

atio

nR

egul

atio

nD

iffer

ence

Diff

eren

cein

corr

uptio

nco

ntro

lind

ex−

0.80

0∗∗∗

−0.

340∗∗∗

−1.

259∗∗∗

−0.

539∗∗∗

0.00

03−

1.07

8∗∗∗

(exp

orte

rmin

usim

port

er)

(0.0

44)

(0.0

65)

(0.0

56)

(0.0

44)

(0.0

64)

(0.0

58)

Diff

eren

cein

asbe

stos

prod

uctio

npe

rcap

ita0.

863∗∗∗

0.37

2∗∗∗

1.35

4∗∗∗

0.56

6∗∗∗

−0.

016

1.14

8∗∗∗

(exp

orte

rmin

usim

port

er)

(0.0

46)

(0.0

68)

(0.0

58)

(0.0

47)

(0.0

68)

(0.0

61)

Diff

eren

cein

envi

ronm

enta

lsus

tain

abili

tyin

dex

0.00

5∗∗∗

0.00

6∗∗∗

0.00

3∗∗∗

(exp

orte

rmin

usim

port

er)

(0.0

002)

(0.0

003)

(0.0

003)

Con

tiguo

us−

0.00

1−

0.00

30.

001

−0.

001

−0.

002

0.00

1(0.0

19)

(0.0

26)

(0.0

20)

(0.0

19)

(0.0

26)

(0.0

20)

Com

mon

lang

uage

−0.

0004

−0.

001

0.00

02−

0.00

03−

0.00

10.

0003

(0.0

09)

(0.0

14)

(0.0

09)

(0.0

09)

(0.0

14)

(0.0

09)

Com

mon

colo

ny0.

0003

0.00

1−

0.00

020.

0002

0.00

1−

0.00

02(0.0

23)

(0.0

34)

(0.0

23)

(0.0

23)

(0.0

33)

(0.0

23)

Log

ofdi

stan

ce−

0.00

1−

0.00

20.

0004

−0.

0004

−0.

001

0.00

05(0.0

04)

(0.0

06)

(0.0

05)

(0.0

04)

(0.0

06)

(0.0

05)

Con

stan

t0.

006

0.01

6−

0.00

40.

005

0.01

4−

0.00

5(0.0

56)

(0.0

89)

(0.0

58)

(0.0

58)

(0.0

92)

(0.0

58)

Mul

tilat

eral

resi

stan

cedu

mm

ies

Yes

Yes

Yes

Yes

Yes

Yes

F-te

stfo

rexc

lude

din

stru

men

ts28

7∗∗∗

38∗∗∗

448∗∗∗

362∗∗∗

163∗∗∗

352∗∗∗

N9,

312

9,31

29,

312

9,31

29,

312

9,31

2

Not

es:S

ingl

e,do

uble

,and

trip

leas

teri

sks

(*,*

*,**

*)in

dica

tesi

gnifi

canc

eat

the

15%

,10%

,and

5%le

vel.

Stan

dard

erro

rs(i

npa

rent

hese

s)ar

ero

bust

tohe

tero

sked

astic

ity.T

hede

pend

entv

aria

bles

are

Exp

orte

rmin

usIm

port

erin

dex

diff

eren

ces.

360 September 2017 Journal of Agricultural and Resource Economics

Tabl

e3.

Est

imat

ing

the

Eff

ecto

fDiff

eren

cesi

nA

ggre

gate

Toba

cco

Reg

ulat

ions

onTo

bacc

oIm

port

Valu

eW

eigh

ted

byE

xpor

ter

and

Impo

rter

GD

PPr

oduc

tUsi

nga

Gra

vity

Equ

atio

nPP

ML

IVPP

ML

Vari

able

(1)

(2)

(3)

(4)

(1)

(2)

(3)

(4)

Con

tiguo

us1.

899∗∗∗

1.95

5∗∗∗

1.91

5∗∗∗

1.97

3∗∗∗

1.81

5∗∗∗

4.67

5∗∗∗

1.95

6∗∗∗

6.83

5∗∗∗

(0.0

77)

(0.0

77)

(0.0

77)

(0.0

77)

(0.4

68)

(1.0

55)

(0.5

46)

(2.5

41)

Com

mon

lang

uage

0.41

5∗∗∗

0.37

9∗∗∗

0.41

2∗∗∗

0.37

2∗∗∗

0.45

7∗1.

081∗∗∗

0.46

6∗3.

418∗∗∗

(0.0

39)

(0.0

39)

(0.0

39)

(0.0

39)

(0.2

87)

(0.4

74)

(0.3

03)

(0.9

40)

Com

mon

colo

ny0.

672∗∗∗

0.63

5∗∗∗

0.61

0∗∗∗

0.55

8∗∗∗

1.57

1∗∗∗

1.37

1∗1.

888∗∗∗

−3.

326∗∗

(0.0

86)

(0.0

86)

(0.0

865)

(0.0

86)

(0.4

57)

(0.9

38)

(0.5

08)

(1.7

67)

Log

ofdi

stan

ce−

0.92

4∗∗∗

−0.

937∗∗∗

−0.

914∗∗∗

−0.

929∗∗∗

−2.

485∗∗∗

−4.

495∗∗∗

−2.

708∗∗∗

−9.

795∗∗∗

(0.0

17)

(0.0

17)

(0.0

17)

(0.0

17)

(0.1

55)

(0.6

44)

(0.2

48)

(1.3

94)

Toba

cco

regu

latio

ndi

ffer

ence

−0.

116∗∗∗

−0.

087∗∗∗

−0.

085∗∗∗

−0.

049

−3.

698∗∗∗

−14.2

53∗∗∗

−7.

454∗∗∗

−24.7

75∗∗∗

(exp

orte

rmin

usim

port

er)

(0.0

39)

(0.0

39)

(0.0

39)

(0.0

39)

(1.1

78)

(3.9

80)

(2.3

25)

(4.3

64)

Toba

cco

regu

latio

ndi

ffer

ence×

0.44

8∗∗∗

0.48

8∗∗∗

11.2

35∗∗∗

41.1

06∗∗∗

bila

tera

lGD

Pdi

ffer

ence

(0.0

39)

(0.0

39)

(3.2

10)

(8.3

54)

Trad

etr

eaty

dum

my

0.45

0∗∗∗

0.53

4∗∗∗

(0.0

48)

(0.0

46)

Con

stan

t−

19.4

75∗∗∗

−20

.069∗∗∗

−19.9

59∗∗∗

−20

.268∗∗∗

−7.

684∗∗∗

13.9

51∗∗∗

−4.

853∗

66.1

02∗∗∗

(0.2

35)

(0.2

06)

(0.2

39)

(0.2

06)

(1.8

11)

(6.8

37)

(2.9

64)

(13.

569)

Mul

tilat

eral

resi

stan

cedu

mm

ies

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

N9,

312

9,31

29,

312

9,31

29,

312

9,31

29,

312

9,31

2

Not

es:S

ingl

e,do

uble

,and

trip

leas

teri

sks

(*,*

*,**

*)in

dica

tesi

gnifi

canc

eat

the

15%

,10%

,and

5%le

vel.

Stan

dard

erro

rs(i

npa

rent

hese

s)ar

ero

bust

tohe

tero

sked

astic

ity.S

peci

ficat

ions

(1)a

nd(2

)ofI

VPP

ML

are

just

iden

tified

,whi

lesp

ecifi

catio

ns(3

)and

(4)a

reov

er-i

dent

ified

.

Galinato, Olanie, and Yoder Trade and Health Effects of Tobacco Regulations 361

Empirical Link between Differences in Tobacco Regulations, Tobacco Trade, and Health

We provide two sets of estimates of first-stage regressions for three endogenous variables (table 2).The first set includes only two instruments—the corruption control index and asbestos production—which lead to just-identified estimates for most of the second-stage specifications. The second setof instruments includes the environmental sustainability index, which overidentifies our second-stage estimates. In most specifications, the instruments are significant and have consistent signs.An increase in the difference between exporter and importer average asbestos production per capitaleads to a larger difference in exporter and importer tobacco regulations. This may be explainedby an increase in perceived marginal damages from tobacco use leading to more demand forstringent regulations and a larger tobacco regulation gap between the bilateral partners. On theother hand, a larger difference in the corruption control index between exporters and importersleads to smaller differences in tobacco regulations between exporters and importers. One potentialexplanation is that when the government can be lobbied to influence regulations, consumer advocacygroups have greater influence on the government than tobacco firms, leading to more stringenttobacco regulations and a larger difference in tobacco regulations. The difference in environmentalsustainability index is also positive and significant, suggesting that countries that implementstringent environmental regulations are also likely to implement more stringent tobacco regulations.All F-stats for excluded instruments are greater than 10, illustrating that our instruments are notweak.

Tables 3 and 4 present coefficient estimates of the determinants of tobacco imports usingthe PPML and IV-PPML models. Standard errors are robust to heteroskedasticity. There are fourspecifications for each type of estimation. Specifications (1) and (2) for the PPML estimates donot include trade treaties, while specifications (3) and (4) include them on top of the multilateralresistance dummies. Specifications (1) and (2) for the IV PPML estimates use the difference inasbestos production and the difference in corruption control as instruments, while specifications (3)and (4) also include differences in the environmental sustainability index.

Our theory suggests that trade costs—in the form of distance between countries, trade barriers,cultural differences, geographic locations, and regulations—affect trade flows. Our empirical resultsshow that the distance effect is negative and significant in most models, suggesting that longerdistances reduce tobacco imports. Having a common language, being in the same colony or tradetreaty, or sharing a border have a positive effect on tobacco imports in most specifications.

The absolute value of the coefficient on the difference in tobacco regulations is smaller in thePPML model than the IV-PPML model. Without accounting for endogeneity of the regulations, weexpect a bias in the PPML estimates relative to the IV-PPML values. One possibility is that a largedifference in tobacco regulations may lead to more tobacco imports between trading pairs. However,more tobacco imports could induce more stringent regulations in the importing country leading toa smaller difference in tobacco regulations and, therefore lower tobacco imports. This may be areason why the PPML estimates are lower in absolute value than the IV-PPML estimates when notcontrolling for endogeneity.

We find support for our hypothesis regarding the effect of difference in aggregate tobaccoregulations between exporters and importers on tobacco import value in the case where the incomegap between exporter and importer is positive and large. From table 3, as the difference betweenaggregate tobacco regulations between exporters and importers increases, tobacco import valuesignificantly decreases. However, when we include an interaction term between differences intobacco regulations and differences in GDP per capita, the magnitude of the coefficient related tothe interaction term is positive and significant. This implies that when trade occurs between a richexporter and poor importer, a larger gap in tobacco regulations leads to more tobacco imports.

362 September 2017 Journal of Agricultural and Resource EconomicsTa

ble

4.E

stim

atin

gth

eE

ffec

tofD

iffer

ence

sin

Dis

aggr

egat

edTo

bacc

oR

egul

atio

nson

Toba

cco

Impo

rtVa

lue

Wei

ghte

dby

Exp

orte

ran

dIm

port

erG

DP

Prod

uctU

sing

aG

ravi

tyE

quat

ion

Vari

able

(1)

(2)

(3)

(4)

(1)

(2)

(3)

(4)

Con

tiguo

us1.

915∗∗∗

1.95

5∗∗∗

1.92

0∗∗∗

1.97

2∗∗∗

2.46

4∗∗∗

2.86

2∗3.

236∗∗∗

2.43

7∗∗

(0.0

77)

(0.0

77)

(0.0

77)

(0.0

77)

(0.8

27)

(1.8

48)

(1.3

25)

(1.3

61)

Com

mon

lang

uage

0.41

0∗∗∗

0.39

0∗∗∗

0.40

8∗∗∗

0.38

1∗∗∗

0.20

9−

0.71

50.

187

−2.

519∗∗

(0.0

39)

(0.0

39)

(0.0

39)

(0.0

39)

(0.3

14)

(2.0

39)

(0.3

54)

(1.4

35)

Com

mon

colo

ny0.

693∗∗∗

0.63

2∗∗∗

0.62

8∗∗∗

0.56

5∗∗∗

1.64

5∗∗∗

1.76

31.

933∗∗∗

3.73

0∗∗

(0.0

86)

(0.0

86)

(0.0

86)

(0.0

86)

(0.5

51)

(1.4

00)

(0.8

64)

(1.9

41)

Log

ofdi

stan

ce−

0.92

2∗∗∗

−0.

939∗∗∗

−0.

917∗∗∗

−0.

932∗∗∗

−2.

331∗∗∗

−4.

775∗∗∗

−2.

270∗∗∗

−7.

288∗∗∗

(0.0

17)

(0.0

17)

(0.0

17)

(0.0

18)

(0.1

88)

(2.2

31)

(0.1

70)

(2.1

52)

Mar

ketin

gan

dco

unte

r-ad

vert

isin

g0.

070∗∗∗

0.08

4∗∗∗

0.07

3∗∗∗

0.09

1∗∗∗

1.49

2−

3.25

03.

732

−4.

159

regu

latio

ndi

ffer

ence

(exp

orte

r–(0.0

30)

(0.0

30)

(0.0

30)

(0.0

30)

(1.9

35)

(3.7

11)

(2.6

61)

(3.3

07)

impo

rter

)

Age

and

smok

ing-

loca

tion

−0.

220∗∗∗

−0.

211∗∗∗

−0.

191∗∗∗

−0.

179∗∗∗

−2.

819∗∗∗

−5.

007∗∗∗

−3.

289∗∗∗

−5.

977∗∗∗

regu

latio

ndi

ffer

ence

(exp

orte

r–(0.0

35)

(0.0

35)

(0.0

35)

(0.0

35)

(0.9

29)

(1.6

35)

(1.0

82)

(2.0

73)

impo

rter

)

Mar

ketin

gan

dco

unte

r-ad

vert

isin

g×

bila

tera

lGD

P0.

056∗∗

0.11

4∗∗∗

12.3

2323

.372∗∗∗

diff

eren

cere

gula

tion

diff

eren

ce(e

xpor

ter–

impo

rter

)(0.0

30)

(0.0

30)

(12.

957)

(7.2

44)

Age

and

smok

ing-

loca

tion

regu

latio

ndi

ffer

ence

0.45

9∗∗∗

0.42

2∗∗∗

−1.

756

−2.

637

(exp

orte

r–im

port

er)×

bila

tera

lGD

Pdi

ffer

ence

(0.0

35)

(0.0

35)

(6.3

54)

(3.4

09)

Trad

eTr

eaty

Dum

my

0.44

1∗∗∗

0.48

5∗∗∗

(0.0

48)

(0.0

46)

Con

stan

t−

20.2

10∗∗∗

−20.1

34∗∗∗

−19.9

35∗∗∗

−20.2

96∗∗∗

−10.0

25∗∗∗

11.1

41−

10.5

63∗∗∗

34.3

75∗∗

(0.2

05)

(0.2

06)

(0.2

39)

(0.2

06)

(2.0

30)

(20.

194)

(1.8

14)

(19.

708)

Mul

tilat

eral

Res

ista

nce

Dum

mie

sY

esY

esY

esY

esY

esY

esY

esY

esN

9,31

29,

312

9,31

29,

312

9,31

29,

312

9,31

29,

312

Not

es:S

ingl

e,do

uble

,and

trip

leas

teri

sks

(*,*

*,**

*)in

dica

tesi

gnifi

canc

eat

the

15%

,10%

,and

5%le

vel.

Stan

dard

erro

rs(i

npa

rent

hese

s)ar

ero

bust

tohe

tero

sked

astic

ity.

Galinato, Olanie, and Yoder Trade and Health Effects of Tobacco Regulations 363

The mean tobacco import value response to differences in tobacco regulations is inelastic and,based on the mean difference in tobacco regulations between unique country pairs, equal to -0.8 fromthe IV-PPML estimates in table 3.12 However, as the income gap increases, the elasticity decreases inabsolute value and eventually becomes positive.13 Based on our estimates, if the difference in GDPper capita between exporter and importer is at least $13,000, the elasticity of tobacco imports givendifferences in tobacco regulations is positive. Approximately 20% of our sample, 1,723 observations,have this income gap.

To illustrate the role of income differences between exporter and importer, consider the UnitedStates, the largest exporter of cigarettes, and Greece, the largest consumer of cigarettes per capita.14

The per capita income difference between the United States and Greece is approximately $25,000,which is associated with an elasticity of tobacco imports due to differences in regulation equal to2.97. If the United States adopts one more marketing regulation, the difference in tobacco regulationindices increases by 0.03, which is a 14.5% increase relative to the mean difference. This impliesthat tobacco imports to Greece from the United States will increase by 43%, which is equal to anannual increase of $0.7 million in tobacco import value.

We disaggregate tobacco regulations into two groups—marketing and counter-advertisingregulations versus age and smoking-location regulations—and summarize their effect on tobaccoimports in table 4. We find that as the difference in counter-advertising and marketing regulatorystringency between exporters and importers of tobacco increases, more tobacco is imported bycountries with less stringent regulations. Furthermore, the effect is magnified when the exportingcountry has a significantly larger income than the importing country in some specifications.The effect is significant with the PPML specifications and when including the differencein environmental sustainability indices between countries as an instrument in the IV-PPMLspecification.

Interestingly, differences in age and smoking-location regulations between exporters andimporters have a consistent negative effect on overall imports. In this case, countries that adoptand enforce similar age restrictions and limits on smoking locations are more likely to engage intobacco trade between one other. Such an effect could be a reflection of the institutional similaritybetween trading partners or may show a need for synchronizing regulations as a precursor for trade,similar to the synchronization found in the EU.

These two results hint at marketing and/or counter-advertising regulations as the main tobaccoregulations that incentivize rich exporters to send tobacco products to poor importers since the effectof age and smoking-location regulations on tobacco imports is negative. A reallocation of advertisingand marketing spending, similar to that shown by Elliott, Wei, and Lenton (2010) from UKtobacco firms, may be the underlying mechanism leading to this result. This hints at a reallocationof advertising spending by tobacco firms away from countries with stringent regulations towardcountries with lax regulations. The policy recommendation for effective tobacco consumptionreduction is to harmonize tobacco regulations across trading partners such that countries with laxregulations adopt more stringent policies similar to their trading partners to mitigate the spillovereffects.

12 The elasticity is calculated as ∂xi j

∂(ri−r j)· (ri−r j)

xi j= (1− σ)α

(ri − r j

), where (1− σ)α =−3.698 in the IV-PPM model

and the mean difference in tobacco regulation for a unique country pair,(ri − r j

), is 0.215.

13 When examining how the elasticity changes with a change in the income gap, we use the results from

specification (2) of the IV PPML in table 3, where the elasticity is now ∂xi j

∂(ri−r j)· (ri−r j)

xi j= (1− σ)α

(ri − r j

)+

(1− σ)β(ri − r j

)(GDPi − GDPj

), where (1− σ)α =−14.253 and (1− σ)β = 11.235.

14 We also conduct PPML for high-income importing countries only, thereby reducing income gaps and tobacco regulationdifferences between bilateral pairs. As with the results from the full sample, we find that the interaction effect of tobaccoregulation differences and GDP differences is positive and significant with magnitudes that are slightly smaller. Results areavailable from the authors upon request.

364 September 2017 Journal of Agricultural and Resource Economics

Table 5. The Elasticity of Tobacco-Related Mortality Across Different Income Gaps BetweenTrading Partners

GDP Per Capita Difference betweenExporter and Importer ($)

Elasticity of Tobacco-Related Mortality withRespect to Tobacco Regulation Difference

-50,000 −0.058∗∗

(0.034)-25,000 −0.035∗∗

(0.021)-15,000 −0.026∗∗

(0.015)-5,000 −0.016∗∗

(0.010)0 −0.012∗∗

(0.007)5,000 −0.007∗∗

(0.004)15,000 0.002

(0.002)25,000 0.011∗∗

(0.007)50,000 0.034∗∗

(0.021)

Notes: Single and double asterisks (*, **) indicate significance at the 15% and 10% level. Theelasticity is derived using the following formula: εM

R = θ Ri/ j × εC

I × εDC × εM

C , where θ Ri/ j is the

elasticity of the value of tobacco imports with respect to regulations by the importer or exporter,εC

I is the elasticity of tobacco import market share from the value of tobacco imports, εDC is the

elasticity of tobacco consumption per capita from the tobacco import market share, and εMC is

the elasticity of mortality from per capita tobacco consumption. We convert the regulationcoefficients in table 2 into elasticities by multiplying them by mean regulations. Based on theexisting literature, we adopt the following values: εM

C = 0.3409 with a standard error of 0.0602(Escario and Molina, 2004), εD

C = 0.031 with a standard error of 0.017, and εCI = 0.366 in the

latest value of their sample in 1995 (Hsieh, Hu, and Lin, 1999). We use the formula for thevariance of a nonlinear univariate function g(A) to calculate the variance of each estimate,

which is equal to V (g(A)) =(

∂g∂A

)TV(A)

(∂g∂A

), where ∂g

∂A is a vector whose ith element is

partial derivative with respect to g and V(A) is the variance-covariance matrix of parameters(Kennedy, 1998). The corresponding standard error (in parentheses) is the square root of thisvariance.

Difference in Tobacco Regulations, Trade, and Mortality

Using our estimates and those in the literature, we simulate the effect that differences in tobaccoregulations have on mortality and morbidity rates in an importing country. We use the estimatesfrom two studies along with our own to obtain an elasticity measure of mortality given a changein the difference in tobacco regulations between country pairs with different income gaps. Escarioand Molina (2004) estimated the effect of per capita tobacco consumption on mortality rates fortobacco-related cancer (including cancer of the lung, trachea, and bronchi) in select EU countries.They found that a 1% increase in per capita tobacco consumption led to a 0.34% increase in totaltobacco-related cancer mortality. Hsieh, Hu, and Lin (1999) estimated the effect of import marketshares of tobacco on tobacco consumption to show that a 1% increase in import market shares led toa 0.031% increase in tobacco consumption per capita. Using their estimates and our own estimatesfrom specification (2) of the IV PPML in table 3, we derive an approximate elasticity of tobacco-related mortality brought about by a change in the difference in tobacco regulations between tradingpartners across different income gaps (table 5).

Galinato, Olanie, and Yoder Trade and Health Effects of Tobacco Regulations 365

Assuming an income gap between exporter and importer of $25,000 (similar to the UnitedStates and Greece), we find that when an exporter adopts more stringent counter-advertisingtobacco regulations, the difference in the aggregate tobacco index increases by 0.25, which leadsto a 1.28% increase in imported tobacco-related cancer mortality in the importing country.15 Theaverage mortality rate from twelve European Union countries due to tobacco-related cancer is 309deaths per million people (Escario and Molina, 2004). The adoption by an exporter of counter-advertising translates to an increase in imported tobacco cancer-related deaths of four smokersper million people annually. Approximately twenty smokers suffer from tobacco-related diseasesfor every one tobacco-related death (Centers for Disease Control and Prevention, 2003). Assumingthis ratio of tobacco-related diseases to deaths, the importing country also experiences an increasein morbidity of eighty smokers per million people annually due to an increase in the exporter’scounter-advertising regulation stringency. The effect is larger as the income gap increase. For a largedeveloping country such as Indonesia, with a population of 242 million, the regulatory effects canbe significant.

As a point of comparison, four females per million people worldwide die annually from alcoholuse disorders (World Health Organization, 2004). Thus, the spillover effects from adopting counter-advertising regulations, all else equal, are comparable to female deaths from disorders from alcoholuse. These effects hint at the potential health benefits of coordinating tobacco regulatory policiesacross countries.

Concluding Remarks

Given that tobacco consumption is projected to increase worldwide (especially in developingcountries), tobacco-related disease is likely to remain a public health policy concern. Although therehas been much analysis regarding the efficacy of tobacco-related regulation in own-country tobaccoconsumption, no one has investigated the difference in tobacco regulations between bilateral tradingpartners and its effect on tobacco trade.

In this article, we empirically estimate the effects of differences in tobacco regulations ontobacco trade. We find these effects to be significant in determining the flow of tobacco imports.There are two striking results with important policy implications. First, tobacco imports increasewhen an exporter’s tobacco regulations are stringent relative to the tobacco regulations of itsimporting trade partner and when the income per capita gap between the exporting country andthe importing country is at least $13,000. Differences in marketing and counter-advertising tobaccoregulations between bilateral trading partners may drive the results, especially if rich exportingcountries target poor importing countries with lax marketing regulations. This may hint at a potential“tobacco disease haven” where developing countries experience a rise in tobacco-related illnessesdue to increased tobacco-regulation stringency in developed countries.

Second, the spillover health effects are significant because when a rich exporting country adoptscounter-advertising tobacco regulations, leading to a larger difference in tobacco regulations, thepoor importing country experiences an increase in mortality and morbidity. When the exporteradopts one additional marketing regulation there is an increase in tobacco-related cancer andmorbidity of four and eighty smokers per million people annually in the poor importing country,respectively, if the income gap between exporter and importer is $25,000. This makes a case forcoordinated increases or harmonization in tobacco regulations to internalize the spillover effect andfurther decrease trade flows of tobacco. Our results also support the idea of lobbying to increasetobacco regulations in developing countries, not only to reduce consumption of domestic tobaccoproducts, as shown in the previous literature, but also to prevent an influx of international tobaccoproducts from developed countries, as shown in this study.

[Received Auguts 2016; final revision received September 2017.]

15 The elasticity of tobacco-related mortality is 0.011 in table 5 when the income gap is 25,000. An increase of 0.25implies a 14.25% increase in the difference in tobacco regulations between exporter and importer.

366 September 2017 Journal of Agricultural and Resource Economics

References

Anderson, J. E., and E. van Wincoop. “Gravity with Gravitas: A Solution to the Border Puzzle.”American Economic Review 93(2003):170–192. doi: 10.1257/000282803321455214.

———. “Trade Costs.” Journal of Economic Literature 42(2004):691–751. doi: 10.1257/0022051042177649.

Baltagi, B., and D. Levin. “Estimating Dynamic Demand for Cigarettes Using Panel Data: TheEffects of Bootlegging, Taxation, and Marketing Reconsidered.” Review of Economics andStatistics 68(1986):148–155.

Blecher, E. “The Impact of Tobacco Advertising Bans on Consumption in Developing Countries.”Journal of Health Economics 27(2008):930–942. doi: 10.1016/j.jhealeco.2008.02.010.

Centers for Disease Control and Prevention. “Cigarette Smoking-Attributable Morbidity — UnitedStates, 2000.” Morbidity and Mortality Weekly Report 52(2003):842–844. Available online athttps://www.cdc.gov/mmwr/pdf/wk/mm5235.pdf.

Chaloupka, F. J., and A. Laixuthai. “U.S. Trade Policy and Cigarette Smoking in Asia.” WorkingPaper 5543, National Bureau of Economic Research, 1996. DOI: 10.3386/w5543.

Chaloupka, F. J., and K. E. Warner. “The Economics of Smoking.” In A. J. Culyer and J. Newhouse,eds., Handbook of Health Economics, vol. 1, Part B. North-Holland, Amsterdam: Elsevier, 2000,1539–1627.

Chapman, S., R. Borland, M. Scollo, R. C. Brownson, A. Dominello, and S. Woodward. “The Impactof Smoke-Free Workplaces on Declining Cigarette Consumption in Australia and the UnitedStates.” American Journal of Public Health 89(1999):1018–1023. doi: 10.2105/AJPH.89.7.1018.

Commission of the European Communities. “Raw Tobacco.” 2003. Available online athttps://ec.europa.eu/agriculture/tobacco/reports/rep_en.pdf.

Disdier, A.-C., L. Fontagné, and M. Mimouni. “The Impact of Regulations on Agricultural Trade:Evidence from the SPS and TBT Agreements.” American Journal of Agricultural Economics90(2008):336–350. doi: 10.1111/j.1467-8276.2007.01127.x.

Elliott, C., Y. Wei, and P. Lenton. “The Effect of Government Policy on Tobacco AdvertisingStrategies.” Bulletin of Economic Research 62(2010):243–258. doi: 10.1111/j.1467-8586.2009.00319.x.

Escario, J. J., and J. A. Molina. “Will a Special Tax on Tobacco Reduce Lung CancerMortality? Evidence for EU Countries.” Applied Economics 36(2004):1717–1722. doi: 10.1080/0003684042000266883.

Feenstra, R. C. Advanced International Trade: Theory and Evidence. Princeton, NJ: PrincetonUniversity Press, 2004.

Felbermayr, G. J., and W. Kohler. “Exploring the Intensive and Extensive Margins of World Trade.”Review of World Economics 142(2006):642–674. doi: 10.1007/s10290-006-0087-3.

Food and Agriculture Organization of the United Nations. Projections of Tobacco Consumption andTrade to the Year 2010. Rome: FAO, 2003. Available online at http://www.fao.org/docrep/006/Y4956E/Y4956E00.HTM.

Gallup, J. L., J. D. Sachs, and A. D. Mellinger. “Geography and Economic Development.”International Regional Science Review 22(1999):179–232. doi: 10.1177/016001799761012334.

Grossman, G. M., and A. B. Krueger. “Environmental Impacts of a North American Free TradeAgreement.” In P. Garber, ed., The US-Mexico Free Trade Agreement, Cambridge, MA: MITPress, 1993.

———. “Economic Growth and the Environment.” Quarterly Journal of Economics110(1995):353–377. doi: 10.2307/2118443.

Helpman, E., M. Melitz, and Y. Rubinstein. “Estimating Trade Flows: Trading Partners andTrading Volumes.” Quarterly Journal of Economics 123(2008):441–487. doi: 10.1162/qjec.2008.123.2.441.

Galinato, Olanie, and Yoder Trade and Health Effects of Tobacco Regulations 367

Holden, C., K. Lee, A. Gilmore, G. Fooks, and N. Wander. “Trade Policy, Health, and CorporateInfluence: British American Tobacco and China’s Accession to the World Trade Organization.”International Journal of Health Services 40(2010):421–441. doi: 10.2190/HS.40.3.c.

Hsieh, C.-R., T.-W. Hu, and C.-F. J. Lin. “The Demand for Cigarettes in Taiwan: Domestic VersusImported Cigarettes.” Contemporary Economic Policy 17(1999):223–234. doi: 10.1111/j.1465-7287.1999.tb00677.x.

Hu, T. W., H. Y. Sung, and T. E. Keeler. “Reducing Cigarette Consumption in California:Tobacco Taxes vs an Anti-Smoking Media Campaign.” American Journal of Public Health85(1995):1218–1222. doi: 10.2105/AJPH.85.9.1218.

Institution for European Environmental Policy. Manual of Environmental Policy. London:Earthscan, 2012. Available online at https://ieep.eu/understanding-the-eu/manual-of-european-environmental-policy.

Kennedy, P. A Guide to Econometrics. Cambridge, MA: MIT Press, 1998.Keohane, N., R. Revesz, and R. Stavins. “The Choice of Regulatory Instruments in Environmental

Policy.” Harvard Environmental Law Review 22(1998):313–367.Knack, S., and P. Keefer. “Institutions and Economic Performance: Cross-Country Tests Using

Alternative Institutional Measures.” Economics & Politics 7(1995):207–227. doi: 10.1111/j.1468-0343.1995.tb00111.x.

Laugesen, M., and C. Meads. “Tobacco Advertising Restrictions, Price, Income and TobaccoConsumption in OECD Countries, 1960âAS1986.” British Journal of Addiction 86(1991):1343–1354. doi: 10.1111/j.1360-0443.1991.tb01710.x.

Levinson, A., and M. S. Taylor. “Unmasking the Pollution Haven Effect.” International EconomicReview 49(2008):223–254. doi: 10.1111/j.1468-2354.2008.00478.x.

Lewit, E. M., D. Coate, and M. Grossman. “The Effects of Government Regulation on TeenageSmoking.” Journal of Law and Economics 24(1981):545–569. doi: 10.1086/466999.

Li, V. C., Y. J. Kim, C. K. Ewart, P. B. Terry, J. C. Cuthie, J. Wood, E. A. Emmett, and S. Permutt.“Effects of Physician Counseling on the Smoking Behavior of Asbestos-Exposed Workers.”Preventive Medicine 13(1984):462–476. doi: 10.1016/0091-7435(84)90015-X.

McFadden, D., J. Wright, B. Wiggs, and A. Churg. “Cigarette Smoke Increases the Penetration ofAsbestos Fibers into Airway Walls.” American Journal of Pathology 123(1986):95–99.

Nelson, J. P. “Cigarette Demand, Structural Change, and Advertising Bans: International Evidence,1970-1995.” Contributions in Economic Analysis & Policy 2(2003):1–29. doi: 10.2202/1538-0645.1111.

Saffer, H., and F. Chaloupka. “The Effect of Tobacco Advertising Bans on Tobacco Consumption.”Journal of Health Economics 19(2000):1117–1137. doi: 10.1016/S0167-6296(00)00054-0.

Santos Silva, J. M. C., and S. Tenreyro. “The Log of Gravity.” Review of Economics & Statistics88(2006):641–658. doi: 10.1162/rest.88.4.641.

———. “On the Existence of the Maximum Likelihood Estimates in Poisson Regression.”Economics Letters 107(2010):310–312. doi: 10.1016/j.econlet.2010.02.020.

———. “Further Simulation Evidence on the Performance of the Poisson Pseudo-Maximum Likelihood Estimator.” Economics Letters 112(2011):220–222. doi: 10.1016/j.econlet.2011.05.008.

Schneider, L., B. Klein, and K. M. Murphy. “Governmental Regulation of Cigarette HealthInformation.” Journal of Law and Economics 24(1981):575–612. doi: 10.1086/467001.

Smith, G. “Gates, Bloomberg Set Up Fund to Help Poor Countries Fight Big Tobacco.” Fortune(2015). Available online at http://fortune.com/2015/03/19/gates-bloomberg-set-up-fund-to-help-poor-countries-fight-big-tobacco/.

Stewart, M. J. “The Effect on Tobacco Consumption of Advertising Bans in OECD Countries.”International Journal of Advertising 12(1993):155–180. doi: 10.1080/02650487.1993.11104533.

Stoforos, C., and G. Mergos. “Modelling EU Tobacco Policy Reform.” Applied Economics36(2004):2167–2175. doi: 10.1080/000368404200068629.

368 September 2017 Journal of Agricultural and Resource Economics

Warner, K. E. “Cigarette Smoking in the 1970s: The Impact of the Antismoking Campaign onConsumption.” Science 211(1981):729–731. doi: 10.1126/science.7455711.

Wooldridge, J. M. Econometric Analysis of Cross-Section and Panel Data. Cambridge, MA: MITPress, 2002.

World Economic Forum. 2001 Environmental Sustainability Index: An Initiative of theGlobal Leaders of Tomorrow Environment Task Force, World Economic Forum. Davos,Switzerland: World Economic Forum, 2001. Available online at http://archive.epi.yale.edu/files/2001_esi_report.pdf.

World Health Organization. “Annex Table 2: Deaths by Cause, Sex and Mortality Stratum inWHO Regions, Estimates for 2002.” In The World Health Report 2004, Geneva: World HealthOrganization, 2004, 120–125. Available online at http://www.who.int/whr/2004/annex/topic/en/annex_2_en.pdf.

———. “WHO Report on the Global Tobacco Epidemic.” 2008. Available online athttp://www.who.int/tobacco/global_report/en/.

Galinato, Olanie, and Yoder Trade and Health Effects of Tobacco Regulations 369

Appendix A: Deriving Import Demand

The first-order conditions that maximize equation (1) subject to the budget constraint are

(σ

σ−1

)[∑

ni=1

(ci jβi

) σ−1σ

] σσ−1−1(

1βi

) σ−1σ (

σ−1σ

)ci j

1σ − pi j = 0 ∀ i = 1... n,(A1)

y j − ∑ni=1 pi jci j = 0,(A2)

where λ is the marginal utility of income. Solving the first-order conditions and defining the nominalvalue of imports to j from i as xi j ≡ pi jci j yields country j’s demand for country i’s goods:

(A3) xi j = y j

(βi pi j

∑ni=1 βi pi j

)1−σ

.

When markets clear, we have the following condition:

(A4) yi =n

∑j=1

y j

(βi piti j

∑ni=1 βi piti j

)1−σ

.

We control for the endogeneity of prices by solving equation (A4) for the scaled prices:

(A5) yi = (βi pi)1−σ

n

∑j=1

y j

(ti j

∑ni=1 βi piti j

)1−σ

,

which when rearranged yields

(A6) βi pi =

yi

(n

∑j=1

y j

(ti j

∑ni=1 βi piti j

)1−σ)−1

11−σ

.

Multiplying the numerator and denominator of equation (A6) by world income, defined as yw ≡∑i yi,yields

(A7) βi pi =

yi

yw

(n

∑j=1

y j

yw

(ti j

∑ni=1 βi piti j

)1−σ)−1

11−σ

.

Substituting equation (A2) into equation (A7) yields

(A8) βi pi =

yi

yw

n

∑j=1

y j

yw

eα(ri−r j)eCijBdbi j (1 + vi j)

∑ni=1 βi pieri j eCijBdb

i j (1 + vi j)

1−σ−1

11−σ

.

Finally, substituting equation (A8) into equation (A3) yields equation (4):

xi j

y jyi=

1yw

eα(ri−r j)eCijBdbi j (1 + vi j)

Pjπi

1−σ

.

370 September 2017 Journal of Agricultural and Resource Economics

Appendix B: Data Sources

Variable Description of Variables SourceTobacco trade value Value of manufactured and

unmanufactured tobacco in thousandsUSD from exporter to importer

World Bank’s COMTRADE dataset

GDP Per capita gross domestic product for2000 in USD

World Bank

Tobacco regulations All tobacco regulation information World Health Organization TobaccoControl Country Profiles, 2000

Contiguous borders Dummy variable indicating whetherthe trading pair share a border

Centre d’Etudes Prospectives etd’Informations Internationales

Common language Dummy variable indicating whetherthe trading pair shares an officiallanguage

Centre d’Etudes Prospectives etd’Informations Internationales

Common colonizer Dummy variable indicating whetherthe trading pair has ever shared acolonial link

Centre d’Etudes Prospectives etd’Informations Internationales

Trade distance A measure of distance between thetrading pair’s largest cities.Calculated using the great circleformula and the cities’ longitude andlatitude coordinates

Centre d’Etudes Prospectives etd’Informations Internationales

Trade Treaty Dummy 1 if countries have a bilateral tradetreaty or belong to the same tradeunion, 0 otherwise

World Trade Organization 2001Annual Report.

Average asbestos production percapita

Average of asbestos production percapita from 1930–1970

Data for asbestos is derived from theU.S. Geological Survey andpopulation data is fromwww.populstat.info

Corruption control index An indicator of corruption within thepolitical system, characterized byfinancial corruption and insidiouscorruption. A score of 0 (lowcorruption control) to 6 (highcorruption control) is assigned

International Country Risk Guidepublished by The PRS Group, Inc.

Environmental Sustainability Index A composite index that evaluates if acountry is on a path that isenvironmental sustainability relativeto other countries using variousmeasures that reflect pollution levels,natural resource endowments andenvironmental regulations. The indexranges from 0 to 100

Yale University’s Center forEnvironmental Law and Policy

Galinato, Olanie, and Yoder Trade and Health Effects of Tobacco Regulations 371

Appendix C: Tobacco Regulation Subcategories

Counter-AdvertisingRegulation Index

MarketingRegulation Index

Age RegulationIndex

Smoking-LocationRegulation Index

Mandated education Marketing in certain media Sales to minors Smoking in governmentbuildings

Mandated health warnings Marketing to certainaudiences

Age verification for sales Smoking in privateworksites

Marketing in certainlocations

Vending machines sales Smoking in educationalfacilities

Sponsorship or promotionfor certain audiences

Free tobacco products Smoking in health carefacilities

Sponsorship marketing ofevents

Smoking on buses

Brand stretching Smoking on trainsMisleading information onpackaging

Smoking in taxis

Package health warning/message

Smoking on ferries

Smoking on domestic airflightsSmoking on internationalflightsSmoking in restaurantsSmoking in nightclubs andbarsSmoking in other publicplaces

Appendix D: Deriving the Determinants of Tobacco Imports Using a Gravity Model and NoImputed Values

PPML IV-PPMLVariable (1) (2) (1) (2)Contiguous 1.737∗∗∗ 1.797∗∗∗ 1.412∗∗∗ 60.671

(0.093) (0.093) (0.517) (52.540)Common language 0.538∗∗∗ 0.503∗∗∗ 0.607∗∗∗ 2.031

(0.045) (0.045) (0.261) (6.158)Common colony −0.198∗∗∗ −0.176∗∗ 0.165 15.665∗

(0.101) (0.101) (0.335) (10.732)Log of distance −0.802∗∗∗ −0.827∗∗∗ −1.548∗∗∗ −32.942

(0.021) (0.021) (0.116) (24.766)Tobacco regulation difference (exporter minusimporter)

−0.632∗∗∗ −0.554∗∗∗ 1.741 −2.008

(0.052) (0.052) (1.349) (8.264)Tobacco regulation difference × bilateral GDPdifference

0.593∗∗∗ 262.370

(0.049) (210.758)Constant −21.073∗∗∗ −20.962∗∗∗ −17.146∗∗∗ 315.899

(0.230) (0.232) (1.416) (270.288)Multilateral resistance dummies Yes Yes Yes YesN 6,006 6,006 6,006 6,006

Notes: Single, double, and triple asterisks (*, **, ***) indicate [statistical] significance at the 15%, 10%, and 5% level. Standard errors (inparentheses) are robust to heteroskedasticity.